otr4j
otr4j copied to clipboard
ibauersachs
Bouncy Castle 1.65 instead of 1.49
Bouncy Castle 1.65 instead of 1.49: https://www.bouncycastle.org/releasenotes.html
http://www.bouncycastle.org/latest_releases.html
This will most likely cause problems:
The default parameter sizes for DH and DSA are now 2048. If you have been relying on key pair generation without passing in parameters generated keys will now be larger. -- BouncyCastle release notes
It will cause protocol violations for otr4j implementations, as they did not hardcode default lengths for DSA signatures.
This will most likely cause problems:
The default parameter sizes for DH and DSA are now 2048. If you have been relying on key pair generation without passing in parameters generated keys will now be larger. -- BouncyCastle release notes
It will cause protocol violations for otr4j implementations, as they did not hardcode default lengths for DSA signatures.
Note: It is since Bouncy Castle 1.57.
We can already done an update from 1.49 to 1.56.
For completeness, I believe I answered you on a different channel already - at least partially:
- True, as of version 1.57 indeed.
- Yes, upgrades up to 1.56 should go alright. (At least, I do not recall running into such nasty issues.)
- Necessary changes in code should be in the
cryptopackage, probably primarilyOtrCryptoEngine(+ -Impl). I seem to recall that SMP-related code wasn't an issue as it does not use DSA, but I'm not completely sure anymore.
@cobratbq Hey :-)
Do we even need BouncyCastle in otr4j? The only relevant usage is in net.java.otr4j.crypto.OtrCryptoEngineImpl and from what I can tell, this is all basic JCE crypto: a bit of AES and DSA.
The problem with key sizes are in implementations of OtrEngineHost. Given that 1024bit keys are insecure anyway, I'd be willing to break those intentionally. See also #34
@cobratbq Hey :-)
Hi!
Do we even need BouncyCastle in otr4j? The only relevant usage is in
net.java.otr4j.crypto.OtrCryptoEngineImpland from what I can tell, this is all basic JCE crypto: a bit of AES and DSA.
I remember there being comments about compatibility issues. I have not checked about removing BC entirely, that could also work.
The problem with key sizes are in implementations of
OtrEngineHost. Given that 1024bit keys are insecure anyway, I'd be willing to break those intentionally. See also #34
Yes, I misremembered that at first. Indeed DSA keypair generation is left as an exercise for the user, in this case Jitsi. The 1024 bit key length - or more precisely the 20 bytes/160 bits public parameter q requirement - is mandated by spec, though. So we cannot deviate from that.
@cobratbq Hey :-)
Hi!
Do we even need BouncyCastle in otr4j? The only relevant usage is in
net.java.otr4j.crypto.OtrCryptoEngineImpland from what I can tell, this is all basic JCE crypto: a bit of AES and DSA.I remember there being comments about compatibility issues. I have not checked about removing BC entirely, that could also work.
Wasn't that some crap about BouncyCastle vs. SpongyCastle (Android)? If it runs on plain Java 8, I see no reason to keep BouncyCastle.
The problem with key sizes are in implementations of
OtrEngineHost. Given that 1024bit keys are insecure anyway, I'd be willing to break those intentionally. See also #34Yes, I misremembered that at first. Indeed DSA keypair generation is left as an exercise for the user, in this case Jitsi. The 1024 bit key length - or more precisely the 20 bytes/160 bits public parameter
qrequirement - is mandated by spec, though. So we cannot deviate from that.
Fine, but then we can set that in KeyPair/KeyFactory.init in Jitsi.
Sure, that will work.
The only reason for keeping BC for this situation is better/simpler API. I'm not sure if we use that. If not, then no reason at all.
Unless JDK complains about 1024 bit DSA keys being weak already. Can't remember from the top of my head.
On Wed, 10 Jun 2020, 01:44 Ingo Bauersachs, [email protected] wrote:
@cobratbq https://github.com/cobratbq Hey :-)
Hi!
Do we even need BouncyCastle in otr4j? The only relevant usage is in net.java.otr4j.crypto.OtrCryptoEngineImpl and from what I can tell, this is all basic JCE crypto: a bit of AES and DSA.
I remember there being comments about compatibility issues. I have not checked about removing BC entirely, that could also work.
Wasn't that some crap about BouncyCastle vs. SpongyCastle (Android)? If it runs on plain Java 8, I see no reason to keep BouncyCastle.
The problem with key sizes are in implementations of OtrEngineHost. Given that 1024bit keys are insecure anyway, I'd be willing to break those intentionally. See also #34 https://github.com/jitsi/otr4j/pull/34
Yes, I misremembered that at first. Indeed DSA keypair generation is left as an exercise for the user, in this case Jitsi. The 1024 bit key length - or more precisely the 20 bytes/160 bits public parameter q requirement - is mandated by spec, though. So we cannot deviate from that.
Fine, but then we can set that in KeyPair/KeyFactory.init in Jitsi.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/jitsi/otr4j/pull/32#issuecomment-641637222, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAOYYVWSNOQEKFICWAPXPMDRV3CNRANCNFSM4MGLT7IQ .
Sure, that will work. The only reason for keeping BC for this situation is better/simpler API.
I already replaced one part of it, what was ~30 lines of BC calls and conversion to regular JCE is now 5 lines, including the try-catch.
I'm not sure if we use that. If not, then no reason at all. Unless JDK complains about 1024 bit DSA keys being weak already. Can't remember from the top of my head.
It probably does for TLS, but not if you call the APIs directly.
Anyway, pointless. I missed #33.